Nvidia Just Made $96 Billion in Three Months, and the AI Spending Race Is Getting Even Bigger

Nvidia has posted another record-breaking quarter. Revenue hit $96.2 billion, its data center business alone generated $89 billion, and Amazon is preparing to deploy another 2 million Nvidia GPUs. The numbers are so large that the AI boom is starting to look less like a technology trend and more like an infrastructure arms race.

Nvidia used to be the company gamers argued about whenever a new graphics card was too expensive.

Now it is selling almost $100 billion worth of technology every three months.

For the quarter ending July 26, Nvidia reported $96.2 billion in revenue, up 106% from the same period last year. Net income reached $59.7 billion, while operating income climbed to $63.7 billion. Those would be enormous numbers for an entire year at many technology companies. For Nvidia, that was one quarter.

The Real Money Is No Longer Gaming

The biggest number in Nvidia's results is $89 billion.

That is how much its Data Center business generated during the quarter, up 117% year over year. Data centers now represent the overwhelming majority of Nvidia's business. So while millions of people still know Nvidia because of GeForce cards, the company's financial engine is now AI.

Its chips are being used to train models, run AI services, power autonomous systems, process enormous datasets and build what Nvidia likes to call AI factories.

CEO Jensen Huang summed up Nvidia's view of the market with a simple line:

"Now, compute is revenue."

The argument is that companies are moving beyond experimenting with AI. They are beginning to use it for real products and services, which means they need more computing power. Conveniently, Nvidia happens to sell quite a lot of computing power.

Amazon Wants Another 2 Million Nvidia GPUs

This might be the best example of how aggressive the AI infrastructure race has become.

AWS and Nvidia announced that Amazon plans to deploy 2 million additional Nvidia GPUs across its global infrastructure during 2027 and 2028. The new capacity will include Blackwell Ultra, Rubin and Rubin Ultra systems.

Two million. Not two million dollars worth of GPUs. Two million actual GPUs.

AWS had already announced plans earlier this year to add more than one million Nvidia GPUs. Nvidia now says demand has exceeded those expectations, leading to another enormous expansion.

The partnership also includes plans for 100,000 GPUs dedicated to US government AI infrastructure, including federal and national security workloads. Most PC builders spend three weeks deciding whether they really need a graphics card upgrade. Amazon looked at a million GPUs and apparently said:

We need more.

Nvidia Thinks $96 Billion Is Just the Beginning

The company is already forecasting approximately $108 billion in revenue for the next quarter, plus or minus 2%. Nvidia says that forecast does not assume any Data Center compute revenue from China. That means Nvidia could cross $100 billion in quarterly sales even without one of the world's largest technology markets contributing to that part of the business. The growth is being driven by companies building increasingly large AI systems and by Nvidia's next generation hardware.

The biggest name there is Vera Rubin.

This Is What a Modern "GPU" Looks Like Now

Skip to around 40:39 for Nvidia's Vera Rubin presentation and 46:51 to see the rack systems.

Calling Vera Rubin a graphics card would be a bit like calling an aircraft carrier a boat.

Nvidia's systems now combine large numbers of GPUs, custom CPUs, high-speed networking and memory inside entire racks designed to operate as one enormous AI computer. Vera Rubin is already in full production, and Nvidia says systems are ramping with partners including Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, CoreWeave and Nebius.

Nvidia has also introduced the Vera CPU, specifically designed for AI workloads and agents. Companies exploring or planning to use Vera include OpenAI, Anthropic, SpaceXAI, Oracle and a long list of cloud and infrastructure providers.

That is an important shift. Nvidia does not want to sell only the GPU inside the AI data center. It increasingly wants to sell the architecture around it too.

AI Is Becoming an Infrastructure Race

The most important part of Nvidia's earnings is not simply that one company made an extraordinary amount of money.

It is where the money is coming from.

Amazon is buying millions of GPUs. Microsoft, Google, Oracle and other cloud providers are building massive AI infrastructure. Governments are investing in national AI systems. Startups and established companies are competing for access to the same computing capacity.

Nvidia also says partnerships with major financial firms could help mobilize more than $500 billion of third-party capital for future AI infrastructure projects, subject to final agreements.

This is no longer just a competition over who has the smartest chatbot. It is becoming a competition over electricity, chips, networking, data centers and the physical machines capable of running increasingly large AI systems.

That also creates some uncomfortable questions. These systems consume enormous amounts of power and require expensive cooling, networking equipment, memory and physical space. Building AI at this scale is becoming as much an engineering and energy problem as a software problem. But right now, demand is still accelerating. Nvidia's latest quarter makes that very difficult to ignore.

$96.2 billion in revenue.

$89 billion from Data Centers.

$59.7 billion in net income.

2 million more GPUs heading to AWS.

And Nvidia is already telling investors to expect roughly $108 billion next quarter. A few years ago, Nvidia was best known for asking gamers whether they wanted ray tracing. Now some of the world's largest companies are asking Nvidia how quickly it can deliver another million GPUs.

That escalated quickly.